【摘要】 针对管道内腐蚀损伤厚度难以定量诊断的问题，提出了一种采用超声导波的管道腐蚀损伤定量成像方法。首先，基于频域有限差分法，建立了管中多路径螺旋传播导波的数值模型，可以在已知厚度图时快速计算出导波接收信号；其次，计算空间内存在随机损伤时的接收信号，重复运行数值模型生成了包含3 500组样本的损伤信号数据库；然后，构建了一维卷积神经网络成像模型，使用生成的数据库训练模型，建立了厚度图和接收信号的映射关系，将接收信号输入成像模型以输出对应的厚度图；最后，通过实验验证了该方法的可行性。实验成像结果与真实值之间的均方差为8.604 8×10^-4，相关性系数为0.711 6，成像模型运行时间为0.538 5 s，这表明该方法可以实现管道中腐蚀损伤的厚度定量成像，且成像效率较高。更多还原

【Abstract】 In response to the challenge of quantitatively diagnosing corrosion damage thickness within pipelines, a quantitative imaging method for pipeline corrosion damage using ultrasonic guided waves is proposed. Firstly, based on the frequency domain finite difference method, a numerical model for multi-path helical propagation of guided waves in pipes is established, enabling rapid calculation of guided wave reception signals when thickness map is known. Secondly, by calculating the received signals in the presence of randomly distributed damage, a database comprising 3 500 samples of damage signals is generated through iteratively running the numerical model. Subsequently, a one-dimensional convolutional neural network imaging model is constructed. The model is trained using the generated database to establish a mapping relationship between thickness maps and reception signals, and inputting the reception signals into the imaging model yields corresponding thickness maps. Finally, the feasibility of the proposed method is experimentally validated. The mean square error between experimental imaging results and actual values is 8.6048×10^-4, the correlation coefficient is 0.711 6, and the imaging model runtime is 0.538 5 seconds. The results indicate that the proposed method can achieve quantitative imaging of corrosion damage thickness within pipelines with high imaging efficiency.更多还原